Alice Group plans to expand its Portable Modular Data Center (PMDC) nationwide using up to 201.1 billion won secured through its September initial public offering (IPO).
This move aims to proactively address infrastructure demand driven by the full-scale adoption of agentic AI. The company recognizes that agentic AI, which connects with more diverse entities than generative AI, will inevitably increase infrastructure needs.
In a recent interview with *Electronic Times*, Kim Jae-won, CEO of Alice Group, stated, “Infrastructure is the core of the AI token economy era. Just as the 2000s internet boom led to a broadband installation surge, we plan to construct and supply 10-megawatt (MW) AI PMDCs at optimal and reasonable prices.”
PMDC is a next-generation data center where core facilities are pre-fabricated in modular form at a factory and assembled on-site like LEGO blocks.
Alice Group adopted AI PMDC due to its ability to efficiently scale AI infrastructure in the shortest timeframe. Even considering legal procedures—such as permits, power grid impact assessments, and community negotiations for new data centers—PMDC remains competitive.
Notably, as data center cooling shifts toward water-based systems, the demand for AI PMDC, which utilizes relatively small spaces under load, is expected to rise. Single-story or two-story designs that cool servers with 40°C hot water without additional cooling systems also maximize energy efficiency.
The current utilization rate of the company's self-owned graphics processing units (GPUs) has reached 100%. Considering this high demand, the company plans to expand its AI PMDC presence nationwide, including in the Honam region. Colocation support for surface or central processing units (CPUs) will be provided only when requested by clients or as necessary.
Ellyse Group's strength also lies in supporting 'Neo Cloud,' a GPU-based platform specialized for AI training and inference workloads. While traditional data centers focused on data storage and access, Neo Cloud offers AI-tailored cloud services, including support for AI training and inference, as well as lightweighting through inference after computation and learning.
Cost efficiency has also been achieved by supporting Ethernet-based clustering instead of InfiniBand. The company confirmed an average transmission speed 2.4 times faster than foreign cloud service providers (CSPs) through its proprietary AI training services. This gives Ellyse Group an edge in cost-performance ratio.
Ellyse Group is also expanding its 'Spot Instance' business, which utilizes idle internal and external GPUs after AI training. The company has increased its GPU-as-a-Service (GPUaaS) offerings for startups, research institutes, and academic institutions, with cumulative cloud customers now reaching 1,200 companies.
The company's five patent applications related to AI PMDC implementation further solidify its competitive edge against latecomers. Ellyse Group aims to sustain and expand its momentum, having already surpassed 34.3 billion won in sales and 10.3 billion won in operating profit in the first half of this year.
CEO Kim stated, “Under our mission to benefit society widely through AI, Ellyse Group will provide the technology and infrastructure that allows everyone to access AI.” He added, “We will also provide education on how to properly utilize AI, supporting everyone in becoming proficient in its use.”
